This thesis is devoted to design a software architecture for Human- Robot Collaboration (HRC), to enhance the robots’ abilities for working alongside humans. We propose FlexHRC, a hierarchical and flexible human-robot cooperation architecture specifically designed to provide collaborative robots with an extended degree of autonomy when supporting human operators in tasks with high-variability. Along with FlexHRC, we have introduced novel techniques appropriate for three interleaved levels, namely perception, representation, and action, each one aimed at addressing specific traits of humanrobot cooperation tasks. The Industry 4.0 paradigm emphasizes the crucial benefits that collaborative robots could bring to the whole production process. In this context, a yet unreached enabling technology is the design of robots able to deal at all levels with humans’ intrinsic variability, which is not only a necessary element to a comfortable working experience for humans but also a precious capability for efficiently dealing with unexpected events. Moreover, a flexible assembly of semi-finished products is one of the expected features of next-generation shop-floor lines. Currently, such flexibility is placed on the shoulders of human operators, who are responsible for product variability, and therefore they are subject to potentially high stress levels and cognitive load when dealing with complex operations. At the same time, operations in the shop-floor are still very structured and well-defined. Collaborative robots have been designed to allow for a transition of such burden from human operators to robots that are flexible enough to support them in high-variability tasks while they unfold. As mentioned before, FlexHRC architecture encompasses three perception, action, and representation levels. The perception level relies on wearable sensors for human action recognition and point cloud data for perceiving the object in the scene. The action level embraces four components, the robot execution manager for decoupling action planning from robot motion planning and mapping the symbolic actions to the robot controller command interface, a task Priority framework to control the robot, a differential equation solver to simulate and evaluate the robot behaviour on-the-fly, and finally a random-based method for the robot path planning. The representation level depends on AND/OR graphs for the representation of and the reasoning upon human-robot cooperation models online, a task manager to plan, adapt, and make decision for the robot behaviors, and a knowledge base in order to store the cooperation and workspace information. We evaluated the FlexHRC functionalities according to the application desired objectives. This evaluation is accompanied with several experiments, namely collaborative screwing task, coordinated transportation of the objects in cluttered environment, collaborative table assembly task, and object positioning tasks. The main contributions of this work are: (i) design and implementation of FlexHRC which enables the functional requirements necessary for the shop-floor assembly application such as task and team level flexibility, scalability, adaptability, and safety just a few to name, (ii) development of the task representation, which integrates a hierarchical AND/OR graph whose online behaviour is formally specified using First Order Logic, (iii) an in-the-loop simulation-based decision making process for the operations of collaborative robots coping with the variability of human operator actions, (iv) the robot adaptation to the human on-the-fly decisions and actions via human action recognition, and (v) the predictable robot behavior to the human user thanks to the task priority based control frame, the introduced path planner, and the natural and intuitive communication of the robot with the human.
|Titolo della tesi:||A Hierarchical Architecture for Flexible Human-Robot Collaboration|
|Data di discussione:||30-apr-2019|
|Appare nelle tipologie:||Tesi di dottorato|